Money-Maker Plan — Cursor

Source: C:\Users\zerou\.cursor\plans\real_money_edge_plan_ed80c0d8.plan.md · Generated 2026-05-11 · Cursor Plan Mode output

Overview

Build a fast-track, evidence-backed program to move each asset class toward Tier-2 live-capital readiness, using current /audit metrics, walk-forward validation, and MySQL/GitHub Actions data health gates. Prioritize immediate edge capture in EQUITY/COMMODITY/ETF while aggressively containing FOREX/overfit risk and closing missing validation coverage.

Current baseline (from live `/audit` data)

Source: audit_dashboard/data/dashboard_data.json::performance.asset_class_health

Class	n	WR %	PF	Verdict
COMMODITY	408	67.4	3.92	strong
EQUITY	443	54.0	1.60	Tier-2 candidate
ETF	100	60.0	1.48	near Tier-2 PF threshold
CRYPTO	7875	47.4	1.39	below Tier-2
FOREX	1825	41.8	0.28	stressed; sizing disabled
BOND	11	54.5	0.66	thin sample

Walk-forward coverage exists only for ETF, CRYPTO, FOREX, EQUITY; COMMODITY + BOND missing.
Drift risk active: hf_stats.concept_drift.drift_alert=true, high KS distance.
Overfit detector flags 12 rows (mostly baby_strats) in fwd_vs_bt_divergence.rows.

Pipeline reality check (GHA + MySQL)

/audit refreshed hourly by .github/workflows/audit-dashboard.yml (cron + push paths + FTP deploy).
Dashboard generation merges local + MySQL non-crypto history via audit_trail/dashboard_generator.py using mysql_fetch_closed_non_crypto(max_age_days=365, limit=5000) from trading_picks.
Read path for /audit performance uses ejaguiar1_stocks (trading_picks) via audit_trail/mysql_client.py.
ejaguiar1_backtests exists mainly for bt_backtest_* writes; not the primary read source yet.

Money-Maker skill status

money-maker-ready is live at .claude/skills/money-maker-ready/SKILL.md.
The specific proposal/report paths cited in the prompt were not found in Cursor's repo snapshot; plan uses the live skill plus current data files as source of truth.

Phase 1 — 72h Triage and Guardrails (no new capital yet)

Run a deterministic money-maker audit report from current payload (freshness + per-class + walk-forward + divergence + drift).
Add hard pre-capital gate: any class with drift_alert=true OR missing walk-forward class block is auto-paper_only regardless of PF/WR.
Create ranked "dragger quarantine candidates" from fwd_vs_bt_divergence.rows and low-PF systems; route through mutation/investigation protocol docs before any blocklist edits.

Phase 2 — Fast-track classes with measurable edge

EQUITY (primary scale candidate)

Target: hold PF≥1.5, WR≥50, n from 443 to 600+ while keeping max drawdown within charter.
Action: increase only top validated EQUITY strategies (by forward PF/WR); cap exposure per docs/PERFORMANCE_CHARTER.md.

COMMODITY (primary alpha candidate)

Target: preserve PF≥2.0 and WR≥55 through walk-forward verification (currently missing by-class walk-forward section).
Action: add commodity class to walk-forward output path in alpha_engine/walkforward_validator.py; verify surfacing in audit_trail/dashboard_generator.py.

ETF (near-threshold candidate)

Target: PF lift from 1.48 to ≥1.5 with n from 100 to 180+; keep consistency ≥80%.
Action: controlled emitter expansion only for ETF strategies with positive OOS decay + acceptable drawdown.

Phase 3 — Contain and repair weak classes

FOREX (stressed)

Keep sizing disabled until PF≥1.2 and WR≥45 for sustained rolling window.
Run dedicated deep-dive report; isolate worst strategy/symbol clusters from divergence + low-PF buckets.

CRYPTO (high-volume, sub-tier)

Do not scale gross capital despite large sample until PF≥1.5 on post-quarantine forward window.
Apply targeted quarantine to overfit baby_strats families flagged in divergence table; do not broad-kill source systems.

BOND (thin sample)

Keep paper-only until n≥100; focus on data generation quality before optimization.

Phase 4 — Database and data-contract upgrades

Add explicit /audit telemetry card for DB lineage: "what % of class stats came from local JSON vs MySQL rows" using audit_trail/dashboard_generator.py.
Add optional backtests-vs-live consistency checks pulling ejaguiar1_backtests aggregates (read-only); surface divergence beside fwd_vs_bt_divergence.
Keep primary live-performance truth on ejaguiar1_stocks.trading_picks until backtests DB schema and sync quality are proven stable.

Success criteria

Within 7 days

All classes shown in walkforward.by_class or explicitly labeled UNVERIFIED in /audit.
Drift alert either cleared or class-level sizing guard active with visible status.
Top 10 draggers have disposition (quarantine, mutate, or monitor) with rationale.

Within 14 days

EQUITY and COMMODITY maintain Tier-2 metrics on rolling forward window.
ETF crosses PF≥1.5 with n≥150.
FOREX PF improves materially from stressed baseline OR remains hard-capped (no silent re-enable).

Real-money ready gate

At least 2 asset classes sustain Tier-2 (PF≥1.5, WR≥50, n≥100, drawdown within charter) for consecutive monitoring windows AND pass drift/divergence checks.